Laser cutting process parameter optimization decision-making system

Abstract

The invention relates to the technical field of laser cutting process parameter optimization, and discloses a laser cutting process parameter optimization decision system which comprises a multi-source data acquisition module, a parameter association network construction module, a core parameter screening module and a dynamic optimization execution module. According to the system, through cooperative work of multiple modules, the problems of low efficiency and large result fluctuation caused by dependence on artificial experience in a traditional method are solved, and meanwhile the analysis capacity of the complex relation between technological parameters and cutting quality is improved. The multi-source data acquisition module integrates real-time and historical data, the parameter association network construction module quantifies parameter interaction strength, the core parameter screening module accurately positions key factors, and the dynamic optimization execution module realizes intelligent adjustment. The cutting quality and efficiency can be remarkably improved, universality and expansibility are achieved, and support is provided for intelligent development of the laser cutting field.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing and advanced processing technology, specifically a laser cutting process parameter optimization decision system. Background Technology

[0002] The field of laser cutting process parameter optimization has long faced challenges, including strong parameter correlations, complex optimization processes, and insufficient consistency in cutting results. In traditional laser cutting, setting process parameters typically relies on operator experience and repeated trials, which is inefficient and yields fluctuating results. Furthermore, due to the non-linear relationship between process parameters and cutting quality, manual adjustments alone are insufficient to fully explore the potential optimization space. Therefore, there is an urgent need for a system capable of integrating multi-source data, analyzing the intrinsic correlations between parameters, and achieving intelligent optimization of process parameters to improve the overall efficiency of laser cutting. Summary of the Invention

[0003] This invention relates to the field of laser cutting process parameter optimization technology, and discloses a deep learning-based laser cutting process parameter optimization decision system. The system includes a multi-source data acquisition module, a parameter association network construction module, a core parameter screening module, and a dynamic optimization execution module. Through the collaborative work of these modules, the system solves the problems of low efficiency and large fluctuations in results caused by the reliance on manual experience in traditional laser cutting process parameter setting, while simultaneously improving the ability to analyze the complex relationship between process parameters and cutting quality.

[0004] The multi-source data acquisition module is responsible for acquiring real-time information and historical records of the laser cutting process from multiple heterogeneous data sources. These data sources include, but are not limited to, laser power sensors, cutting speed monitoring equipment, gas pressure regulators, and defocusing detection devices.

[0005] This module performs preliminary cleaning of the collected data, removing noise and outliers, and transforms the unstructured data into a standard format suitable for subsequent analysis. The standard-formatted data is stored in a unified data storage unit, providing the basic input for subsequent modules.

[0006] The parameter association network construction module extracts multi-dimensional process parameter feature vectors from the standardized data storage units. These feature vectors cover the specific values ​​and trends of key parameters such as laser power, cutting speed, gas pressure, and defocusing amount.

[0007] By calculating the interaction strength between feature vectors, this module constructs a one-dimensional parameter correlation graph, where each node represents a process parameter and the edge weights indicate the degree of mutual influence between parameters. Furthermore, the one-dimensional parameter correlation graph generates a global parameter correlation network through a multi-layer fusion algorithm, comprehensively reflecting the complex relationships between various process parameters.

[0008] The core parameter selection module is based on a global parameter association network and combines the distribution characteristics of multi-dimensional process parameter feature vectors to calculate the dispersion index of each parameter. The dispersion index is used to measure the range of variation and stability of a parameter under different cutting conditions.

[0009] Simultaneously, this module introduces a correlation strength index to evaluate the interaction strength between each parameter and other parameters. By comprehensively analyzing the dispersion index and the correlation strength index, a set of core parameters that have a significant impact on cutting quality is selected.

[0010] The dynamic optimization execution module, based on a set of core parameters, drives the dynamic adjustment of laser cutting process parameters. This module trains a deep learning model on historical cutting data to establish a mapping relationship between process parameters and cutting quality.

[0011] During the actual cutting process, this module receives real-time feedback information on the current process parameters and adjusts the specific values ​​of parameters such as laser power, cutting speed, gas pressure, and defocusing amount based on the prediction results of the deep learning model. The adjusted parameters are directly transmitted to the laser cutting equipment through the control interface, completing the real-time update of the parameters.

[0012] The aforementioned modules achieve complete system functionality through the organic integration of hardware and software. The multi-source data acquisition module connects to various sensors and monitoring devices via industrial communication protocols, ensuring the real-time nature and accuracy of data acquisition. The parameter correlation network construction module and the core parameter filtering module run on a high-performance computing platform, utilizing a distributed computing framework to process large-scale datasets. The dynamic optimization execution module connects to the laser cutting equipment through an embedded control system, enabling the rapid issuance and execution of parameter adjustment commands.

[0013] Compared with the prior art, the beneficial effects of the present invention are:

[0014] This invention solves the problem of relying on manual experience in the optimization of traditional laser cutting process parameters through a multi-module collaborative design. The introduction of a multi-source data acquisition module allows the system to fully utilize real-time and historical data, avoiding the limitations of a single data source. The parameter correlation network construction module reveals the intrinsic relationships between process parameters by quantifying the interaction strength, providing a scientific basis for subsequent optimization. The core parameter selection module accurately identifies key factors affecting cutting quality through comprehensive analysis of parameter dispersion and correlation strength. The dynamic optimization execution module, based on the predictive capabilities of a deep learning model, enables intelligent adjustment of process parameters, significantly improving cutting quality and efficiency.

[0015] This invention also possesses strong versatility and scalability. The multi-source data acquisition module supports the access of various types of sensors, adapting to the needs of different models of laser cutting equipment. The parameter association network construction module adopts a scalable algorithm architecture, capable of continuously optimizing the parameter association model as the data scale grows. The dispersion and association strength indices of the core parameter screening module can be flexibly adjusted according to specific application scenarios, meeting the requirements of diverse cutting tasks. The dynamic optimization execution module, through modular design, facilitates the integration of new optimization algorithms and control strategies, further enhancing the system's intelligence level.

[0016] This invention proposes a novel laser cutting process parameter optimization decision-making system through the organic integration of multiple modules. This system not only solves the problems of low efficiency and large result fluctuations in traditional methods, but also utilizes deep learning technology to uncover the complex relationship between process parameters and cutting quality, providing strong support for the intelligent development of the laser cutting field. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall architecture of the laser cutting process parameter optimization decision system of the present invention.

[0018] Figure 2 This is a flowchart of the multi-source data acquisition module.

[0019] Figure 3 This is a diagram illustrating the operational principle of the parameter association network construction module.

[0020] Figure 4 Functional structure diagram of the core parameter filtering module.

[0021] Figure 5 The operation flowchart for dynamically optimizing the execution module.

[0022] Figure 6 This is a schematic diagram showing the connection between the system of the present invention and the laser cutting equipment. Detailed Implementation

[0023] This invention provides a deep learning-based laser cutting process parameter optimization decision system, the overall architecture of which is as follows: Figure 1 As shown in the attached diagram, the system comprises a multi-source data acquisition module, a parameter correlation network construction module, a core parameter filtering module, and a dynamic optimization execution module. These modules achieve the system's complete functionality through the organic integration of hardware and software, and are connected to the laser cutting equipment via an industrial communication protocol and an embedded control system. The specific implementation methods of each module will be described in detail below with reference to the accompanying drawings.

[0024] As a preferred implementation, the multi-source data acquisition module is the basic data input part of the system, and its workflow is as follows: Figure 2As shown, this module acquires real-time information and historical records of the laser cutting process from multiple heterogeneous data sources, including but not limited to laser power sensors, cutting speed monitoring devices, gas pressure regulators, and defocusing detection devices.

[0025] In practical applications, these data sources are connected to the multi-source data acquisition module via industrial communication protocols to ensure the real-time nature and accuracy of data acquisition. The acquired data first undergoes preliminary cleaning to remove noise and outliers; for example, a moving average algorithm is used to smooth sensor data and eliminate random fluctuations. Subsequently, the unstructured data is converted to a standard format and stored in the data storage unit. The data storage unit employs a distributed database architecture, supporting efficient storage and rapid retrieval of large-scale data, providing the foundational input for subsequent modules.

[0026] The location relationships of the multi-source data acquisition module are clearly defined. Its input end is directly connected to various sensors and monitoring equipment, while its output end is connected to the parameter association network construction module.

[0027] As a preferred implementation, the operating principle of the parameter association network construction module is as follows: Figure 3 As shown, this module extracts multi-dimensional process parameter feature vectors from the data storage unit. These feature vectors cover the specific values ​​and trends of key parameters such as laser power, cutting speed, gas pressure, and defocusing amount.

[0028] In practice, the parameter correlation network construction module first calculates the interaction strength between feature vectors, quantifying the degree of mutual influence between different parameters through correlation analysis algorithms. For example, the correlation between laser power and cutting speed may exhibit a non-linear relationship, which is modeled using a Gaussian kernel function. Subsequently, a one-dimensional parameter correlation graph is generated, where each node represents a process parameter, and the edge weights represent the degree of mutual influence between parameters.

[0029] As a preferred implementation, the one-dimensional parameter correlation graph generates a global parameter correlation network through a multi-layer fusion algorithm. This network comprehensively reflects the complex relationships between various process parameters. The parameter correlation network construction module is located after the multi-source data acquisition module. Its input is connected to the data storage unit, and its output provides the global parameter correlation network as input to the core parameter filtering module.

[0030] As a preferred implementation, the functional structure of the core parameter filtering module is as follows: Figure 4As shown, this module is based on a global parameter correlation network and combines the distribution characteristics of multi-dimensional process parameter feature vectors to calculate the dispersion index and correlation strength index of each parameter. The dispersion index is used to measure the range of variation and stability of a parameter under different cutting conditions. For example, variance analysis can be used to evaluate the fluctuation of gas pressure under different material thicknesses.

[0031] The correlation strength index assesses the interaction strength between each parameter and other parameters; for example, it quantifies the coupling degree between laser power and cutting speed using a mutual information algorithm. In practical applications, the core parameter screening module uses a comprehensive analysis of the dispersion index and the correlation strength index to select a set of core parameters that have a significant impact on cutting quality.

[0032] For example, in stainless steel cutting tasks, laser power and defocusing amount may be identified as core parameters, while the influence of gas pressure is relatively small. The core parameter screening module has a clear positional relationship; its input receives the global parameter association network from the parameter association network construction module, and its output passes the selected set of core parameters to the dynamic optimization execution module.

[0033] As a preferred implementation method, the operation flow of the dynamically optimized execution module is as follows: Figure 5 As shown, this module, based on a set of core parameters, drives the dynamic adjustment of laser cutting process parameters. In practical applications, the dynamic optimization execution module trains historical cutting data using a deep learning model to establish a mapping relationship between process parameters and cutting quality.

[0034] For example, a convolutional neural network is used to model the relationship between the surface roughness of the cut surface and parameters such as laser power and cutting speed. During the actual cutting process, the dynamic optimization execution module receives feedback information on the current process parameters in real time and adjusts the specific values ​​of parameters such as laser power, cutting speed, gas pressure, and defocusing amount based on the prediction results of the deep learning model.

[0035] The adjusted parameters are directly transmitted to the laser cutting equipment via the control interface, enabling real-time parameter updates. The dynamic optimization execution module has a clearly defined position; its input receives the core parameter set from the core parameter filtering module, and its output is connected to the laser cutting equipment via the control interface.

[0036] As a preferred embodiment, the connection method between the system of the present invention and the laser cutting equipment is as follows: Figure 6 As shown, the multi-source data acquisition module connects to various sensors and monitoring devices via industrial communication protocols to ensure the real-time performance and accuracy of data acquisition. The parameter correlation network construction module and the core parameter filtering module run on a high-performance computing platform, utilizing a distributed computing framework to process large-scale datasets.

[0037] The dynamic optimization execution module is connected to the laser cutting equipment through an embedded control system, enabling the rapid issuance and execution of parameter adjustment commands.

[0038] In practical applications, such as for cutting aluminum alloy sheets, the system first acquires real-time data through a multi-source data acquisition module, then analyzes the interrelationships between parameters through a parameter association network construction module, and the core parameter screening module selects laser power and cutting speed as core parameters. Finally, the dynamic optimization execution module dynamically adjusts the parameter settings based on the prediction results of the deep learning model, thereby optimizing the cutting quality.

[0039] This invention solves the problem of relying on manual experience in the optimization of traditional laser cutting process parameters by organically integrating the above modules. The multi-source data acquisition module supports the access of various types of sensors, adapting to the needs of different laser cutting equipment models. The parameter association network construction module adopts a scalable algorithm architecture, which can continuously optimize the parameter association model as the data scale grows. The dispersion and association strength indices of the core parameter screening module can be flexibly adjusted according to specific application scenarios to meet the requirements of diverse cutting tasks. The dynamic optimization execution module, through modular design, facilitates the integration of new optimization algorithms and control strategies, further improving the system's intelligence level.

[0040] Example 1: This example is for the cutting of 304 stainless steel plates with a thickness of 5-10mm. The laser cutting process parameter optimization decision system is used to achieve a synergistic improvement in cutting quality and efficiency. The cutting quality is reflected in the perpendicularity of the cut and the surface roughness.

[0041] Operation of the multi-source data acquisition module:

[0042] Data source connection: The multi-source data acquisition module is connected to the laser power sensor, cutting speed encoder, nitrogen pressure sensor, and defocusing laser displacement meter via the ModbusTCP industrial communication protocol to acquire the following data in real time:

[0043] Laser power: 1500-2500W, sampling interval 0.1s;

[0044] Cutting speed: 2-6 m / min, dynamic variation recorded;

[0045] Nitrogen pressure: 0.8-1.2 MPa, fluctuation range ±0.05 MPa;

[0046] Defocusing amount: -1mm to +1mm, the trajectory is dynamically adjusted according to the thickness of the sheet material.

[0047] Data preprocessing: The raw data is smoothed using a moving average algorithm to remove high-frequency noise from the sensors, such as outliers like the instantaneous jump of laser power by ±50W; unstructured cutting process logs (such as "cutting speed suddenly drops at 10:05") are converted into standardized JSON format and stored in a Hadoop distributed database, supporting 1000 data entries per second.

[0048] Running the parameter association network building module:

[0049] Feature vector extraction: Multi-dimensional features are extracted from standardized data, including average laser power, cutting speed fluctuation rate, peak gas pressure, cumulative change in defocusing amount, etc., to form a 128-dimensional feature vector.

[0050] One-dimensional association graph generation: The interaction strength between feature vectors is calculated using the Pearson correlation coefficient, for example:

[0051] There is a negative correlation between laser power and cutting speed, with a correlation coefficient of -0.72: the speed needs to be reduced when the power increases to avoid ablation.

[0052] There is a positive correlation between gas pressure and decoking amount, with a correlation coefficient of 0.65: when the decoking amount increases, the pressure needs to be increased to remove the slag.

[0053] Using nodes to represent parameters and edge weights to represent interaction strength, ranging from 0 to 1, four one-dimensional parameter association graphs are generated.

[0054] Global network fusion: A multi-layer fusion algorithm (based on graph neural network GNN) is used to integrate the single-dimensional graph into a global parameter association network, which intuitively presents the triangular relationship closed loop of "laser power-cutting speed-gas pressure", with a total weight of 68%.

[0055] Operation of the core parameter filtering module:

[0056] Dispersion index calculation: Parameter volatility is assessed through analysis of variance.

[0057] Laser power variance: 12000W², with significant fluctuations under different plate thicknesses;

[0058] Defocus variance: 0.05mm², indicating high stability.

[0059] Association strength index calculation: Quantifying parameter coupling degree through mutual information algorithm:

[0060] Mutual information value between laser power and surface roughness: 0.83 (strong correlation);

[0061] Mutual information value between gas pressure and cut perpendicularity: 0.76 (strong correlation).

[0062] Core parameters were determined by combining dispersion (weight 40%) and correlation strength (weight 60%), and laser power (overall score 0.89) and gas pressure (overall score 0.82) were selected as core parameters.

[0063] Dynamically optimize the execution of the module:

[0064] Model training: A convolutional neural network was used to train the model on 5,000 sets of historical data (input: laser power, gas pressure; output: surface roughness Ra), and the model accuracy reached 92%.

[0065] Real-time adjustment logic:

[0066] When the surface roughness Ra > 3.2 μm, the model predicts that the laser power needs to be increased by 5%, such as from 2000W to 2100W, while the gas pressure needs to be reduced by 3%, such as from 1.0MPa to 0.97MPa.

[0067] When the verticality deviation of the cut is greater than 0.5°, adjust the gas pressure by +2% to maintain the laser power.

[0068] Results: After parameter adjustment, the surface roughness of stainless steel sheet cutting decreased from an average of 4.5μm to 2.8μm, the pass rate increased from 78% to 95%, and the cutting efficiency increased by 12%.

[0069] Example 2: Optimization of laser cutting process parameters for thin-walled aluminum alloy parts. This example is for thin-walled 6061 aluminum alloy parts with a thickness of 1-3mm, which are prone to deformation and slag adhesion. The system achieves a balance between low deformation and high speed.

[0070] Operation of the multi-source data acquisition module:

[0071] Data source: Connected to laser power sensor, cutting speed monitor, oxygen pressure regulator, and defocusing detection device; data collected includes:

[0072] Laser power: 800-1500W, focused on the material surface;

[0073] Cutting speed: 5-8m / min, rapid cutting is required to avoid thermal deformation;

[0074] Oxygen pressure: 0.4-0.6 MPa, for auxiliary combustion and slag removal.

[0075] Data processing: Instantaneous fluctuations in oxygen pressure are filtered through a moving average algorithm, and standardized data is stored in a distributed database, supporting fast retrieval by the dimension of "material thickness-cutting speed".

[0076] Running the parameter association network building module:

[0077] Feature vector: Extract 86-dimensional features such as laser power peak value, cutting speed acceleration, oxygen pressure integral value, defocusing amount stabilization time, etc.

[0078] Correlation analysis: The one-dimensional correlation graph shows that the interaction intensity of "cutting speed-laser power" is 0.81, indicating that high-speed cutting requires high power to avoid slag adhesion. The interaction intensity of "oxygen pressure-deformation" is 0.77, indicating that insufficient pressure can easily lead to slag residue and deformation.

[0079] Global Network: A global network with "cutting speed" as the core node is generated through a multi-layer fusion algorithm. The edge weights of the network with laser power and oxygen pressure are 0.81 and 0.77, respectively.

[0080] Operation of the core parameter filtering module:

[0081] Dispersion: The cutting speed variance reaches 3.2 m² / min², which is the largest fluctuation in thin-walled part cutting; the laser power variance is 1500 W², and the stability is moderate.

[0082] Correlation strength: The mutual information value between cutting speed and deformation is 0.89 (strong correlation), and the mutual information value between oxygen pressure and slag adhesion rate is 0.85 (strong correlation).

[0083] Key parameters: Cutting speed (overall score 0.91) and oxygen pressure (overall score 0.87) were selected.

[0084] Dynamically optimize the execution of the module:

[0085] Model training: The model was trained using a convolutional neural network on 3000 sets of historical data (input: cutting speed, oxygen pressure; output: deformation δ), with a model error of <0.02mm.

[0086] Real-time adjustments:

[0087] When the detected deformation δ > 0.1 mm, reduce the cutting speed by 8%, such as from 7 m / min to 6.44 m / min; increase the oxygen pressure by 5%, such as from 0.5 MPa to 0.525 MPa.

[0088] When the slag adhesion rate is greater than 3%, increase the oxygen pressure by 10% while maintaining the cutting speed unchanged.

[0089] Results: The deformation of thin-walled aluminum alloy parts decreased from an average of 0.15 mm to 0.08 mm, the slag adhesion rate decreased from 5% to 1.2%, and the cutting efficiency remained at 6.5 m / min, meeting the production cycle requirements.

[0090] Both embodiments demonstrate the system's ability to accurately select core parameters and dynamically optimize processes based on different material properties through multi-module collaboration. Compared with traditional manual experience-based adjustments, the stability of cutting quality is improved by more than 20%, and efficiency is improved by 10%-15%.

[0091] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0092] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A laser cutting process parameter optimization decision system, characterized in that, It includes a multi-source data acquisition module, a parameter correlation network construction module, a core parameter filtering module, and a dynamic optimization execution module; The multi-source data acquisition module is used to acquire real-time information and historical records of the laser cutting process from multiple heterogeneous data sources, and to convert unstructured data into a standard format and store it in the data storage unit. The parameter association network construction module is used to extract multi-dimensional process parameter feature vectors from the standardized data storage unit and generate a global parameter association network. The core parameter filtering module is used to calculate the dispersion index and the correlation strength index based on the global parameter association network to filter out the core parameter set. The dynamic optimization execution module is used to drive the dynamic adjustment of laser cutting process parameters based on the core parameter set.

2. The laser cutting process parameter optimization decision system as described in claim 1, characterized in that, The multi-source data acquisition module is connected to the laser power sensor, cutting speed monitoring equipment, gas pressure regulator, and defocus detection device via an industrial communication protocol. The acquired data includes the specific values ​​and trends of laser power, cutting speed, gas pressure, and defocus.

3. The laser cutting process parameter optimization decision system as described in claim 1, characterized in that, The parameter association network construction module generates a one-dimensional parameter association graph by calculating the interaction strength between the feature vectors of multi-dimensional process parameters, and transforms the one-dimensional parameter association graph into a global parameter association network through a multi-layer fusion algorithm.

4. The laser cutting process parameter optimization decision system as described in claim 1, characterized in that, The core parameter screening module calculates the dispersion index of each parameter through variance analysis and calculates the correlation strength index between each parameter and other parameters through mutual information algorithm.

5. The laser cutting process parameter optimization decision system as described in claim 1, characterized in that, The dynamic optimization execution module trains historical cutting data using a deep learning model to establish a mapping relationship between process parameters and cutting quality, and adjusts the specific values ​​of laser power, cutting speed, gas pressure, and defocusing amount based on feedback information from the current process parameters.

6. The laser cutting process parameter optimization decision system as described in claim 5, characterized in that, The deep learning model uses a convolutional neural network to model the relationship between the surface roughness of the cut surface and parameters such as laser power and cutting speed.

7. The laser cutting process parameter optimization decision system as described in claim 1, characterized in that, The multi-source data acquisition module uses a moving average algorithm to smooth sensor data in order to remove noise and outliers.

8. The laser cutting process parameter optimization decision system as described in claim 1, characterized in that, The data storage unit adopts a distributed database architecture, which supports efficient storage and fast retrieval of large-scale data.

9. The laser cutting process parameter optimization decision system as described in claim 1, characterized in that, The parameter association network building module runs on a high-performance computing platform and uses a distributed computing framework to process large-scale datasets.

10. The laser cutting process parameter optimization decision system as described in claim 1, characterized in that, The dynamic optimization execution module is connected to the laser cutting equipment through a control interface, and the adjusted process parameters are transmitted to the laser cutting equipment through the control interface.

Images